TITLE:
The Prediction of Severe Acute Pancreatitis: From Traditional Scoring Systems to Emerging Artificial Intelligence
AUTHORS:
Hao Li, Chuanxin Zou, Lingyan Hu
KEYWORDS:
Severe Acute Pancreatitis, Prediction, Scoring Systems, Biomarkers, CT Severity Index, Artificial Intelligence, Machine Learning, XGBoost, Random Forest
JOURNAL NAME:
Journal of Biosciences and Medicines,
Vol.14 No.8,
August
12,
2026
ABSTRACT: Severe acute pancreatitis (SAP) is a severe complication of acute pancreatitis (AP), with a mortality rate approaching 30%, bringing a heavy burden to the global healthcare system. Early identification of high-risk patients within the first 48 hours of admission is crucial for timely implementing intensive care, reasonably triaging patients, and carrying out targeted treatments. In the past, various scoring systems were mainly used in clinical practice to evaluate disease severity, including the Ranson score, Glasgow/Imrie score, Acute Physiology and Chronic Health Evaluation II (APACHE II), and BISAP score. Although the above scoring tools can stratify patients by risk, there are issues such as delayed calculation time and suboptimal sensitivity and specificity. Apart from this, although serological indicators such as blood urea nitrogen, procalcitonin, C-reactive protein, and interleukins can provide certain prognostic reference value, the effectiveness of assessing each indicator individually varies significantly. In addition, imaging assessment methods represented by the Balthazar CT severity index can provide detailed information on organ morphology, but this examination requires a delayed scanning time to obtain optimal imaging results, which also limits its clinical application value. In recent years, a variety of advanced machine learning algorithms—including Extreme Gradient Boosting, Random Forests, Support Vector Machines, and Artificial Neural Networks—have developed rapidly. These models can simultaneously integrate patients’ demographic information, serological test results, and imaging data, and they demonstrate consistently stable and excellent performance in overall prognosis prediction, with AUC values generally exceeding 0.90. Therefore, this paper systematically reviews the development of predictive models for severe acute pancreatitis, from traditional scoring systems to emerging machine learning models, summarizing the advantages and disadvantages of each method one by one, while also discussing the challenges that still need to be overcome before such machine learning tools can be effectively applied in clinical practice.